The RT Related Games, Benchmarks, Software, Etc Thread.

[nvidiamd]

sadly the motion in df video is not the motion i talk about

cutscenes, things moving on the screen do not count. dlss can still handle them. or slow movements for that matter

what matters is when you move the camera around, move around, and see if dlss can keep or not. chances are it wont look "sharp" as native 1800p in movement. this, i'm sure of. but DF shills would never expose that


native 1800p should resolve detail in movement so much better

Yes sadly this may be the problem with any temporal solution. They can look great while you stand still but look bad when in motion because you need more frames to get a crisp image and you may not get them while moving. The problem is you won't be bothered too much with the decreased quality when moving this is why we have techs like radeon boost or VRS. So we might have good enough images when you stand still which mean nothing because you won't sit still while gaming ( but it helps to promote the tech ) and not that good images while moving but it is fine because you won't notice too much degradation anyway. Well they are upscalers we shouldn't expect miracles from any of them.
At least a tech like FSR is more honest, the quality is pretty bad all the time.

 

[Nexus18]

The main and only con of dlss has been addressed and now suddenly we have issues in other areas of motion to do with dlss.... which don't seem to have been brought up ever before????

DF talk extensively about motion with dlss here too:

https://youtu.be/dtbqJXb1UDw?t=748

And show all the flaws of the ghosting/motion issues with cyberpunk etc., most of which are ones even I never noticed such as the elevator bit so not sure how exactly they can be "shills" given they highlight issues that a lot of people probably never even noticed before....

Did you watch linus's video linked above too for their blind test?

Maybe my eyesight has gotten worse as I get older but motion/ghosting is my biggest issue when watching or playing any content, it's why I prefer my 60HZ OLED over my 144 IPS monitor as nothing can compete to oled 0.00001ms pixel response and the way it handles motion. If gaming on my monitor, for good motion, FPS ideally needs to be at least 90/100.

Also, might be worth experimenting yourself to see how FPS affects motion clarity:

https://www.testufo.com/framerates-versus

At the end of the day, if you don't like dlss (or even FSR) for whatever reason, no one is forcing you to use it, Linus summed it up perfectly in his video, we have these upscaling methods because we simply don't have the hardware here and now to be able to enjoy next gen titles at high res. or/and high fps. and won't for a good while yet, dlss/fsr allows us to enjoy future like tech. and much improved graphics at playable res. + fps right now.

And if we are referring to motion being pin sharp in movement (especially "fast" movement) then you aren't going to get that even with dlss/fsr off.... you'll be wanting 240+hz lcd monitor AND pushing 200+ fps too..... or a good CRT or oled display.

 

[oguzsoso]

bro, you really don't get it

we're not talking about sample and hold blur lcd screens produce
we don't talk about motion blur that happens regardless due to human eye limitation
we don't talk about ghosting in motion
we don't talk about NONE OF those. stop RESPONDING to me IF you' can't COMPREHEND what is BEING TOLD.


https://imgsli.com/NTk1MzM/1/2

THIS is what we talk about. DLSS CAN resolve detail and make the image look SHARPER when you stand still. look at how sharp textures look. once you move the camera, move your character forward and backwards and see around, YOU will see that RESOLUTION of the image TAKES a big hit.

I REALLY CAN'T COMPREHEND MYSELF why YOU CAN'T seem to UNDERSTAND OR COMPREHEND.

Here's WHAT DF says:

DLSS with INTERNAL 1080P looks sharper than consoles' NATIVE 1800P image.

THEY ARE SKEWING their IDEAS with STANDING STILL SHOTS. DLSS OR NOT, 1080P INPUT RENDER RESOLUTION CANNOT LOOK MORE DETAILED/SHARPER THAN NATIVE 1800P IN MOVEMENT. in ACTUAL movement. ACTUAL movement where YOU MOVE YOUR DAMN CAMERA AROUND. RUN AROUND. NOT ASTEROIDS PASSING BY CALMLY. NOT WHERE THE DOOMGUY SLOWLY WALKS IN A CORRIDOR. GET IT?

IN MOVEMENT. WHEN YOU MOVE YOUR CAMERA LIKE YOU NORMALLY DO WHILE PLAYING. FAST PACE. NOT THINGS MOVING AROUND IN A SMALL PACE IN THE SCREEN.

GET IT? CAPICHE?

anyone with a 4k screen can observe the same effect on 4k DLSS PERFORMANCE while MOVING and STANDING STILL

https://imgsli.com/NTk1MzM/3/0

MOVING... VERSUS STANDING STILL, NOT MUCH HAPPENING ON THE SCREEN, CAPICHE?

Without DLSS, NATIVE 4K WILL LOOK SHARP EVEN IN MOVEMENT, BECAUSE IT CAN RESOLVE MORE DETAIL DUE TO HAVING HIGHER PIXEL COUNT? GET it? YOU DO UNDERSTAND NOW? IS SHOUTING EVEN WORKS?

For the sake of being calm and civil, I will block you again. You really CAN'T comprehend stuff. HAVE fun with your blurry moving image. keep having the delusion that DLSS is perfect and gives better image quality at 1080p render compared to consoles' 1800p native render.

FOR THE SAKE OF Gods. JUST SHOOT UP A GAME. STAND STILL FOR 5-6 SECONDS. SEE HOW THE IMAGE LOOKS SHARP AND GORGEUS.

NOW, TURN AROUND AND TURN BACK TO WHERE YOU WAS LOOKING. QUICKLY GLANCE WHAT KIND OF HORRIBLE BLURRY MESS IT BECOMES. AND STOP BLURTING OUT NONSENSE.

 

[oguzsoso]

Yes sadly this may be the problem with any temporal solution. They can look great while you stand still but look bad when in motion because you need more frames to get a crisp image and you may not get them while moving. The problem is you won't be bothered too much with the decreased quality when moving this is why we have techs like radeon boost or VRS. So we might have good enough images when you stand still which mean nothing because you won't sit still while gaming ( but it helps to promote the tech ) and not that good images while moving but it is fine because you won't notice too much degradation anyway. Well they are upscalers we shouldn't expect miracles from any of them.
At least a tech like FSR is more honest, the quality is pretty bad all the time.

they bother me, more than bothering me, what ACTUALLY bothers me is the DISHONESTY of digital foundry when it comes to DLSS.

1800p native resolution will resolve more detail and will stay MORE crisp compared to dlss input 1080p resolution.

when standing still, 4k dlss performance will look as if NATIVE 4k. because that's how it operates.

they literally SCAM people into believing that DLSS with 1080p input looks better than consoles' native 1800p rendering at 4K, in reality, in the majority of the time you play, 1800p native render will look much crisp and better. they TRICK people into believing that PC is superior with stupid standing still comparisons. that's what bothers me. i don't expect any MIRACLES out of DLSS or FSR. BUT. BUT digital FOUNDRY makes it seem so like DLSS is a miracle. and look at all those people in the comment, saying how CONSOLE is inferior even when rendering 1800p. they literally direct people's thoughts

freaking 500 dollar console handling 1800p native 60 fps ray tracing and these dudes gloat about PC master race people by saying even rtx 2060 PROVIDES better image quality due to DLSS. all the while showing standing still shots. stationary scenes. how, how biased you have gotta be to make such claims?

if any DIGITAL foundry representative happens by my comments, I dare you to COMPARE making a 180 DEGREE turn with doomslayer and compare the IMAGE sharpness in movement with dlss performance at 4k and native 1800p. I DOUBLE dare you. make it %400 zoom. SHOW how FLAWED DLSS ACTUALLY IS, when you actually PLAY the goddamn game

 

[Nexus18]

So you're talking about motion in moving scenarios.... shock horror, when items are in motion (including real life), you lose quality, yes, dlss might make it ever so worse but are you going to notice when in "motion".... nope.

Again have a look at this site to see how motion works and educate yourself on the topic as you are completely clueless as shown time and time again, prime example being trying to run cyberpunk maxed at 4k with a 3070 and blaming your 3070 vram for the fps drops.....



PS. I don't really see anything wrong with that comparison images above, not to mention it is an incredibly flawed test, come back to this when you have a proper test/comparison and have educated yourself some more please.

Might want to take a chill pill too bud, seem to be losing your nerve with all the caps



Too many people are making a mountain out of a molehill and not educated enough on the topic at hand as per usual it seems.

 

[Aegis]

@oguzsoso that Metro shot you linked to is using DLSS 1.x - not the best comparison. 2.x is *much* improved.

*Edit* here - this is a grab I did at 1440p using DLSS 2.2.9 (Quality) whilst waving the mouse around like a maniac (I turned off VSync and motion blur).

https://www.dropbox.com/s/zrfy0e6drcjoik2/Doom_DLSS_1440p.png?dl=1

 

[oguzsoso]

@oguzsoso that Metro shot you linked to is using DLSS 1.x - not the best comparison. 2.x is *much* improved.

*Edit* here - this is a grab I did at 1440p using DLSS 2.2.9 (Quality) whilst waving the mouse around like a maniac (I turned off VSync and motion blur).

https://www.dropbox.com/s/zrfy0e6drcjoik2/Doom_DLSS_1440p.png?dl=1

no, its 2.2, enhanced version

thank you for proving my point, the image you posted looks blurry, not even close to native 1440p

stand still and wait for 4-5 seconds. see how the image looks what it is supposed to look like.

 

[Nexus18]

@oguzsoso that Metro shot you linked to is using DLSS 1.x - not the best comparison. 2.x is *much* improved.

*Edit* here - this is a grab I did at 1440p using DLSS 2.2.9 (Quality) whilst waving the mouse around like a maniac (I turned off VSync and motion blur).

https://www.dropbox.com/s/zrfy0e6drcjoik2/Doom_DLSS_1440p.png?dl=1

I did the same too a while back as wanted to get a comparison with dlss 2.2.9 but sadly with the gamepass version you can't change any files. I noticed some bad ghosting/trailing with these windows (always find dlss 2.1 falls hard with windows) but it doesn't really show up well in these screenshots (these are taken with mouse moving like mad too), first 2 images show it better i.e. the window on the left:

Spoiler

Yuck.... Looks like 720p, maybe even 480p.....

/s

And some more pics where there is "no difference" in ray traced reflections

Spoiler

Next lot is the classic example of SSR reflections where if you look down or up i.e. basically any angle that isn't directing focusing on the reflection, they just become blurred/smeary....

 

[Aegis]

thank you for proving my point, the image you posted looks blurry, not even close to native 1440p

Time to get some new glasses - zoom into the barrel at 400% - there's super-sampling on all the poly edges and a visible dither pattern (I suspect to alleviate 8-bit banding) but there's zero blur. Check out the pin-sharp two-pixel spires to the right of the gun... whatever you're seeing, it ain't DLSS.

 

[oguzsoso]

no lol you don't get it do you

just stand still. wait for image to stabilize. move around. see for yourself

if you can't see difference between standing still and moving image quality, then you better ger yourselves some glass

oh and obligatory note: you're straight going into ignored list.

you people are simply nvidia-dlss shills. its very funny to see how defensive you get for a tech that makes the image blurry while moving the camera. crazy. i have no time dealing with such folk. sorry, just move on along and happily game with your blurry as hell moving image. try playing native 1440p, see how it actually resolves details without dlss. but of course, you wouldn't even try that, otherwise you would be spoiled and not turn back to dlss. i know what boils your blood , lol

you dont even need native 1440p. just stand still for 5 secs with DLSS on and see how the image gets more clear and sharper. if you cant do that, you're simply afraid of TRUTH.

 

[nvidiamd]

Time to get some new glasses - zoom into the barrel at 400% - there's super-sampling on all the poly edges and a visible dither pattern (I suspect to alleviate 8-bit banding) but there's zero blur. Check out the pin-sharp two-pixel spires to the right of the gun... whatever you're seeing, it ain't DLSS.

But it is easy for you to make a comparison slide: take the first screenshot while turning around and then you look at the picture find the same place in the game and you take the second screenshot while you stand still.

 

[Nexus18]

Popcorn worthy this Absolutely mind blowing how clueless people are, perhaps console is the best way forward for some....

 

[oguzsoso]

But it is easy for you to make a comparison slide: take the first screenshot while turning around and then you look at the picture find the same place in the game and you take the second screenshot while you stand still.

they don't even want to understand the point, all I say is DLSS looks superior when standing still and less detailed when panning the camera or moving to a different place. you give it 4-5 secs to staiblize and all of a sudden image gets "native like". this is what I protest, if DF uses this to claim "dlss is so superior, it beats 1800p console rendering", I will get offended. Easy as that

IF dlss could present the same IQ when moving the camera, THEN it would look better than consoles. but THAt's not the case

 

[Wrinkly]

they don't even want to understand the point, all I say is DLSS looks superior when standing still and less detailed when panning the camera or moving to a different place. you give it 4-5 secs to staiblize and all of a sudden image gets "native like". this is what I protest, if DF uses this to claim "dlss is so superior, it beats 1800p console rendering", I will get offended. Easy as that

IF dlss could present the same IQ when moving the camera, THEN it would look better than consoles. but THAt's not the case

There is something wrong with your setup if it is taking 4-5 seconds to stabilise. It should take no more than a few frames in the worst conditions and be masked somewhat by both pixel and human eyesight response. Can you get an uncompressed 1440p video?

 

[Nexus18]

Getting "offended" over a well detailed technical analysis comparison..... Oh lordy

 

[oguzsoso]

i gave it as a good measure, sometimes it can take up to 1-2 seconds. i didn't make the calculations. i already proved how blurry the game becomes in metro exodus once you move your camera, and someone tried to counter it as claiming it was dlss 1.0 (he couldn't accept the fact so he just straight lied up about the situation). that's why i also ignored him, to make his own points true, he wents to the lenghts of saying it was dlss 1.0, in reality it was dlss 2.2 with updated dll. this is how much people are biased towards dlss, it seems like they've forgotten how native image actually resolves the detail when moving the camera

it doesn't matter, as long as you move with your character forward, everything becomes blurry and low res. no amount of discussion can change this HARD fact/truth. i proveed it, guy tried to claim it was dlss 1.0, its not, discussions is closed

i don't really see why so many people are choosing to be this much oblivious to truth

once nvidia hardware is strong enough to render things without DLSS, i'm pretty sure most of these people will start ******** on DLSS, i'm %100 sure

they don't even want to understand that I have no TROUBLES with it making the image lower res when moving. IT is normal. it uses a LOWER res as baseline.

the DISCUSSION TOPIC was about DIGITAL FOUNDRY claiming dlss 4k performance looking better than native 1800p, WHICH is a HUGE LIE. they don't want to accept this. dsr 4K dlss performance will ONLY look better than native 1800p when you STAND STILL. once you moving, that sharpness and "native like" image quality will be gone.

 

[Aegis]

Ok, so here's a magnified crop of both native 1440p and DLSS Quality with my waggling the mouse like crazy (making it difficult to do an exact apples-to-apples comparison): https://www.dropbox.com/s/pwjdcpygsntyl4s/Doom_Comparison.png?dl=1

But honestly, nit-picking the quality of the moving image for DLSS is pointless because, as you can see from the attached image, Doom streams textures whilst the camera is in motion hence why both images look 'blurry'.

Enjoy

*Edit* and just to prove I'm not cheating, here's both versions with the camera stationary (note the improved texture detail in both images): https://www.dropbox.com/s/s0fwp2lvpfaf6e7/Doom_Comparison_Static.png?dl=1

Just for good measure, a waggly mouse Metro Exodus grab (unfortunately a JPEG): https://www.dropbox.com/s/qdr57gegyg1h3mk/Metro_DLSS.jpg?dl=1

This one does show small DLSS artifacts (if you zoom in close on the gloves you can see them) - I can't see any issues with motion on the trees though like in oguzsoso's images. Textures are maybe a bit meh (see on the left) but again, that's more likely down to the engine than DLSS (which can only work with what it's given).

 

[stooeh]

How do we know DLSS is the culprit?

Here's a pic taken while fast panning the camera:

And here's one where I have tried to take a stationary shot from the same place:


Looks sharper? or is it just the variable lighting?

DLSS is off.

If you can get a clip of the image sharpening after you stop moving the camera I would like to see it as I have been unable to replicate it.

 

[Aegis]

How do we know DLSS is the culprit?

As I mentioned above, it could be texture streaming (culling higher detail textures as you're panning around) and the GI's accumulation cache is constantly rebuilding itself in Metro as you move around so that's going to change up the lighting solution.

Ironically for oguzsoso (since he's blocked me ) he's actually right - stating that DLSS can reconstruct an image superior to native 4k from a 1080p base is somewhat disingenuous - in the Doom grabs I took, I could clearly see (when zoomed in) where 960p edges and thin details had been scaled to a higher resolution (they're just not as defined, and some missing details DLSS seems unable to resolve) - that's probably better working with a higher base resolution though (i.e. 1080p -> 4k).

It's certainly got lots of room for improvement, but with 2.2.6+ I'm hard pressed to see why anyone that can benefit from it wouldn't use it. It's hands-down the best reconstruction technique we have so far.

 

[nvidiamd]

In theory, while moving, you will get lower quality image ( a lot of blur ) even on native if it is using TAA, for the same reason you will get lower image quality when using DLSS. But for something like DLSS perf, the image degradation should be bigger because it is not doing only aliasing but also upscaling for a much lower res.